The CEOs and founders of five wireless LAN switch start-ups broke bread with Network World editors at Rosemary's Restaurant during NetWorld+Interop, Las Vegas. The idea was that between bites of Ahi tuna and goat cheese hazelnut cheesecake we would chew on the issues of enterprise WLANs and this new class of product called wireless switches.
Not surprisingly, all of the executives - from Airespace Inc., AirFlow Networks Inc., Aruba Wireless Networks Inc., Trapeze Networks Inc. and Vivato Inc. - are convinced that WLANs are poised to create a huge new market, with plenty of room at the table for ambitious new entrants, namely themselves.
The Wi-Fi wireless Ethernet movement in some ways looks like the early days of the first Web browser, Mosaic, said Airespace CEO Brett Galloway, who earlier in his career served as director of engineering for wireless pioneer Metricom. People are getting their hands on Wi-Fi, finding it easy to use, and building applications for it that are attracting more devices, he said.
"You put that kind of market-adoption cycle together with the enormous untapped power of Wi-Fi . . . at some point, it's an unstoppable force," he said.
But all five agreed that unstoppable force is being restrained by the current architecture of Wi-Fi: individual access points connecting wirelessly to client devices and by cable to wiring closet switches.
"Think about access points," said Robert Machlin, CEO of AirFlow Networks and a veteran of network companies such as Ascend and Cascade. "They were built for (public-access) hot spots and the small office/home office market. But if you want to go to an enterprise deployment, where you use more than one access point, then how are you going to tie these together from a management viewpoint? What we're saying around this table is there's a systems approach to this problem."
That approach features a box, with Layer 2 and sometimes Layer 3 switching features, that plugs smoothly into existing wired networks. But it also has a wealth of added software to identify, track, secure and manage users logging in through a Wi-Fi radio connection. All five companies offer this basic solution, although one of them, Vivato, uses phased-array antennas to form and direct Wi-Fi beams at specific client devices.
Although conceptually simple, these boxes will lead to some big changes in enterprise networks, the CEOs agreed. "We call it 'wireless LAN switching' mostly because it is an add-on to the existing LANs, at some level," says Keerti Melkote, co-founder of Aruba Wireless Networks.
"But we're radically changing what a LAN switch is, by radically changing what a wired LAN switch understands as a user," he said.
In traditional networks, switches don't need to know who a user is, because the cable that connects the switch to a PC anchored on a desk, and uniquely identified with a media access control or IP address, assumes that the PC user is the correct user.
"With wireless, you need to know who that guy really is before you let him onto the network," said Melkote, a one-time member of Intel's IT staff. "And that fundamental departure leads to a lot of different capabilities that you need to integrate into the switch."
"A good term for that is a 'user-aware network' instead of a 'device-aware network,'" Machlin said. "Take (established equipment makers such as) Extreme, Foundry and Cisco, do they just add support for wireless users (to their boxes)? And is that good enough?"
"If you look at a traditional wired switch, it actually doesn't do very much," Airespace's Galloway said. "The curious thing about wireless is that it's mobile. There are huge security issues as a result. You have to create a whole security framework that's a lot harder to do than what you have in an Ethernet switch. And you have the wireless (radio) medium itself, which doesn't operate by itself (as Ethernet does)."
The Vivato difference
Vivato CEO Ken Biba argued that these comments stressed technology too much over the "value proposition" to enterprise users. "The core argument is what does it cost?" he said. "What does it cost to deploy? What is the cost of Wi-Fi coverage? What is the cost of a megabit per square meter? Those finances ultimately drive the success of Wi-Fi."
Biba cited the example of a well-known Wall Street brokerage firm, with whose network executives he had spoken recently. "We had a wonderful discussion about wireless LAN security, and we said to them, 'Golly, your guys are ready to deploy. What's your deployment plan?' And they said, 'Oh, we don't have any money to do that,'" he said.
Not surprisingly, Biba said Vivato's phased-array technology, in a one-yard-wide panel that combines access point and switch, cuts deployment costs dramatically because customers don't need to design, build and run WLANs consisting of scores of separate access points. But his rivals countered that they can exploit the same hardware trends that have made computers relatively inexpensive.
Trapeze CEO Jim Flach insisted that wireless switches can handle a whole range of wireless LAN problems, whether or not they make use of special antennas.
"Our approach is to leverage the commoditization that's already happening around the access point," Aruba's Melkote said. "There's a bunch of vendors already doing radio chipsets. They cost a hundred bucks today. That same chipset is going to cost 10 bucks soon. Whether it's deploying five of those, or 10 or 20, no one will care (because they're so inexpensive)."
Vendors and users face a trio of variables in WLAN calculations, Galloway suggested. "You have a triangle: cost, (radio) coverage and (network) capacity," he said. "And how you optimize those depends on circumstances. If you have an environment that's very obstructed for radio waves, then using a pico-cellular architecture (of numerous access points) can be very cost-effective. In other environments, it's going to make sense to create a coverage and capacity model that is more around managing the radio frequency beams."
Taking on incumbents
All the CEOs said the network heavyweights don't have the solutions corporations need. "The traditional switch makers, like Extreme, don't even know how to spell 'RF,'" Machlin said.
Flach derided Cisco's approach of moving its IOS systems software into its line of Aironet access points. "If you treat the access point as a managed element in your network, you enormously increase your costs," said Flach, a former venture capitalist who has served as CEO or chairman of roughly a dozen network companies. By contrast, he said, Trapeze treats the access point as a stripped-down component, little more than a radio, that the switch can monitor and configure.
Cisco was clearly on the minds of all the participants, which between them have raised US$74.5 million in venture funding. Melkote said two of the most common questions asked by investors and others are: "Why can't Cisco just put this wireless LAN switch technology in a switch blade?" And "When is Cisco going to buy you?"
All five agreed that putting voice traffic over WLANs was sparking keen interest among their corporate prospects, although few were actually doing so at present. Melkote said that small branch offices are proliferating, with attendant high costs to provision them. Voice converged with data over a common wireless IP LAN could make such provisioning less expensive, he said.
Galloway said the WLAN is actually the "killer application" for voice over IP (VoIP). "One of the things that's been missing for VoIP is a driver for you to put in lots of additional ports, and buy lots of additional handsets," he said. "If you're just doing incremental additions, you're going to stick with the current architecture of PBXs and so on." But with wireless LANs, VoIP suddenly makes sense, he insisted.
Even though N+I was much smaller than usual, it did nothing to dampen the enthusiasm of the executives.
"This is the only area of networking with life," AirFlow's Machlin said.